xref: /openbmc/linux/fs/f2fs/node.h (revision c6ac4c0ec416e77cab09cac6cee2d100fbd7fc82)
1 /*
2  * fs/f2fs/node.h
3  *
4  * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5  *             http://www.samsung.com/
6  *
7  * This program is free software; you can redistribute it and/or modify
8  * it under the terms of the GNU General Public License version 2 as
9  * published by the Free Software Foundation.
10  */
11 /* start node id of a node block dedicated to the given node id */
12 #define	START_NID(nid) ((nid / NAT_ENTRY_PER_BLOCK) * NAT_ENTRY_PER_BLOCK)
13 
14 /* node block offset on the NAT area dedicated to the given start node id */
15 #define	NAT_BLOCK_OFFSET(start_nid) (start_nid / NAT_ENTRY_PER_BLOCK)
16 
17 /* # of pages to perform readahead before building free nids */
18 #define FREE_NID_PAGES 4
19 
20 /* maximum readahead size for node during getting data blocks */
21 #define MAX_RA_NODE		128
22 
23 /* control the memory footprint threshold (10MB per 1GB ram) */
24 #define DEF_RAM_THRESHOLD	10
25 
26 /* vector size for gang look-up from nat cache that consists of radix tree */
27 #define NATVEC_SIZE	64
28 
29 /* return value for read_node_page */
30 #define LOCKED_PAGE	1
31 
32 /*
33  * For node information
34  */
35 struct node_info {
36 	nid_t nid;		/* node id */
37 	nid_t ino;		/* inode number of the node's owner */
38 	block_t	blk_addr;	/* block address of the node */
39 	unsigned char version;	/* version of the node */
40 };
41 
42 enum {
43 	IS_CHECKPOINTED,	/* is it checkpointed before? */
44 	HAS_FSYNCED_INODE,	/* is the inode fsynced before? */
45 	HAS_LAST_FSYNC,		/* has the latest node fsync mark? */
46 	IS_DIRTY,		/* this nat entry is dirty? */
47 };
48 
49 struct nat_entry {
50 	struct list_head list;	/* for clean or dirty nat list */
51 	unsigned char flag;	/* for node information bits */
52 	struct node_info ni;	/* in-memory node information */
53 };
54 
55 #define nat_get_nid(nat)		(nat->ni.nid)
56 #define nat_set_nid(nat, n)		(nat->ni.nid = n)
57 #define nat_get_blkaddr(nat)		(nat->ni.blk_addr)
58 #define nat_set_blkaddr(nat, b)		(nat->ni.blk_addr = b)
59 #define nat_get_ino(nat)		(nat->ni.ino)
60 #define nat_set_ino(nat, i)		(nat->ni.ino = i)
61 #define nat_get_version(nat)		(nat->ni.version)
62 #define nat_set_version(nat, v)		(nat->ni.version = v)
63 
64 #define inc_node_version(version)	(++version)
65 
66 static inline void set_nat_flag(struct nat_entry *ne,
67 				unsigned int type, bool set)
68 {
69 	unsigned char mask = 0x01 << type;
70 	if (set)
71 		ne->flag |= mask;
72 	else
73 		ne->flag &= ~mask;
74 }
75 
76 static inline bool get_nat_flag(struct nat_entry *ne, unsigned int type)
77 {
78 	unsigned char mask = 0x01 << type;
79 	return ne->flag & mask;
80 }
81 
82 static inline void nat_reset_flag(struct nat_entry *ne)
83 {
84 	/* these states can be set only after checkpoint was done */
85 	set_nat_flag(ne, IS_CHECKPOINTED, true);
86 	set_nat_flag(ne, HAS_FSYNCED_INODE, false);
87 	set_nat_flag(ne, HAS_LAST_FSYNC, true);
88 }
89 
90 static inline void node_info_from_raw_nat(struct node_info *ni,
91 						struct f2fs_nat_entry *raw_ne)
92 {
93 	ni->ino = le32_to_cpu(raw_ne->ino);
94 	ni->blk_addr = le32_to_cpu(raw_ne->block_addr);
95 	ni->version = raw_ne->version;
96 }
97 
98 static inline void raw_nat_from_node_info(struct f2fs_nat_entry *raw_ne,
99 						struct node_info *ni)
100 {
101 	raw_ne->ino = cpu_to_le32(ni->ino);
102 	raw_ne->block_addr = cpu_to_le32(ni->blk_addr);
103 	raw_ne->version = ni->version;
104 }
105 
106 enum mem_type {
107 	FREE_NIDS,	/* indicates the free nid list */
108 	NAT_ENTRIES,	/* indicates the cached nat entry */
109 	DIRTY_DENTS	/* indicates dirty dentry pages */
110 };
111 
112 struct nat_entry_set {
113 	struct list_head set_list;	/* link with other nat sets */
114 	struct list_head entry_list;	/* link with dirty nat entries */
115 	nid_t set;			/* set number*/
116 	unsigned int entry_cnt;		/* the # of nat entries in set */
117 };
118 
119 /*
120  * For free nid mangement
121  */
122 enum nid_state {
123 	NID_NEW,	/* newly added to free nid list */
124 	NID_ALLOC	/* it is allocated */
125 };
126 
127 struct free_nid {
128 	struct list_head list;	/* for free node id list */
129 	nid_t nid;		/* node id */
130 	int state;		/* in use or not: NID_NEW or NID_ALLOC */
131 };
132 
133 static inline void next_free_nid(struct f2fs_sb_info *sbi, nid_t *nid)
134 {
135 	struct f2fs_nm_info *nm_i = NM_I(sbi);
136 	struct free_nid *fnid;
137 
138 	spin_lock(&nm_i->free_nid_list_lock);
139 	if (nm_i->fcnt <= 0) {
140 		spin_unlock(&nm_i->free_nid_list_lock);
141 		return;
142 	}
143 	fnid = list_entry(nm_i->free_nid_list.next, struct free_nid, list);
144 	*nid = fnid->nid;
145 	spin_unlock(&nm_i->free_nid_list_lock);
146 }
147 
148 /*
149  * inline functions
150  */
151 static inline void get_nat_bitmap(struct f2fs_sb_info *sbi, void *addr)
152 {
153 	struct f2fs_nm_info *nm_i = NM_I(sbi);
154 	memcpy(addr, nm_i->nat_bitmap, nm_i->bitmap_size);
155 }
156 
157 static inline pgoff_t current_nat_addr(struct f2fs_sb_info *sbi, nid_t start)
158 {
159 	struct f2fs_nm_info *nm_i = NM_I(sbi);
160 	pgoff_t block_off;
161 	pgoff_t block_addr;
162 	int seg_off;
163 
164 	block_off = NAT_BLOCK_OFFSET(start);
165 	seg_off = block_off >> sbi->log_blocks_per_seg;
166 
167 	block_addr = (pgoff_t)(nm_i->nat_blkaddr +
168 		(seg_off << sbi->log_blocks_per_seg << 1) +
169 		(block_off & ((1 << sbi->log_blocks_per_seg) - 1)));
170 
171 	if (f2fs_test_bit(block_off, nm_i->nat_bitmap))
172 		block_addr += sbi->blocks_per_seg;
173 
174 	return block_addr;
175 }
176 
177 static inline pgoff_t next_nat_addr(struct f2fs_sb_info *sbi,
178 						pgoff_t block_addr)
179 {
180 	struct f2fs_nm_info *nm_i = NM_I(sbi);
181 
182 	block_addr -= nm_i->nat_blkaddr;
183 	if ((block_addr >> sbi->log_blocks_per_seg) % 2)
184 		block_addr -= sbi->blocks_per_seg;
185 	else
186 		block_addr += sbi->blocks_per_seg;
187 
188 	return block_addr + nm_i->nat_blkaddr;
189 }
190 
191 static inline void set_to_next_nat(struct f2fs_nm_info *nm_i, nid_t start_nid)
192 {
193 	unsigned int block_off = NAT_BLOCK_OFFSET(start_nid);
194 
195 	f2fs_change_bit(block_off, nm_i->nat_bitmap);
196 }
197 
198 static inline void fill_node_footer(struct page *page, nid_t nid,
199 				nid_t ino, unsigned int ofs, bool reset)
200 {
201 	struct f2fs_node *rn = F2FS_NODE(page);
202 	if (reset)
203 		memset(rn, 0, sizeof(*rn));
204 	rn->footer.nid = cpu_to_le32(nid);
205 	rn->footer.ino = cpu_to_le32(ino);
206 	rn->footer.flag = cpu_to_le32(ofs << OFFSET_BIT_SHIFT);
207 }
208 
209 static inline void copy_node_footer(struct page *dst, struct page *src)
210 {
211 	struct f2fs_node *src_rn = F2FS_NODE(src);
212 	struct f2fs_node *dst_rn = F2FS_NODE(dst);
213 	memcpy(&dst_rn->footer, &src_rn->footer, sizeof(struct node_footer));
214 }
215 
216 static inline void fill_node_footer_blkaddr(struct page *page, block_t blkaddr)
217 {
218 	struct f2fs_checkpoint *ckpt = F2FS_CKPT(F2FS_P_SB(page));
219 	struct f2fs_node *rn = F2FS_NODE(page);
220 
221 	rn->footer.cp_ver = ckpt->checkpoint_ver;
222 	rn->footer.next_blkaddr = cpu_to_le32(blkaddr);
223 }
224 
225 static inline nid_t ino_of_node(struct page *node_page)
226 {
227 	struct f2fs_node *rn = F2FS_NODE(node_page);
228 	return le32_to_cpu(rn->footer.ino);
229 }
230 
231 static inline nid_t nid_of_node(struct page *node_page)
232 {
233 	struct f2fs_node *rn = F2FS_NODE(node_page);
234 	return le32_to_cpu(rn->footer.nid);
235 }
236 
237 static inline unsigned int ofs_of_node(struct page *node_page)
238 {
239 	struct f2fs_node *rn = F2FS_NODE(node_page);
240 	unsigned flag = le32_to_cpu(rn->footer.flag);
241 	return flag >> OFFSET_BIT_SHIFT;
242 }
243 
244 static inline unsigned long long cpver_of_node(struct page *node_page)
245 {
246 	struct f2fs_node *rn = F2FS_NODE(node_page);
247 	return le64_to_cpu(rn->footer.cp_ver);
248 }
249 
250 static inline block_t next_blkaddr_of_node(struct page *node_page)
251 {
252 	struct f2fs_node *rn = F2FS_NODE(node_page);
253 	return le32_to_cpu(rn->footer.next_blkaddr);
254 }
255 
256 /*
257  * f2fs assigns the following node offsets described as (num).
258  * N = NIDS_PER_BLOCK
259  *
260  *  Inode block (0)
261  *    |- direct node (1)
262  *    |- direct node (2)
263  *    |- indirect node (3)
264  *    |            `- direct node (4 => 4 + N - 1)
265  *    |- indirect node (4 + N)
266  *    |            `- direct node (5 + N => 5 + 2N - 1)
267  *    `- double indirect node (5 + 2N)
268  *                 `- indirect node (6 + 2N)
269  *                       `- direct node
270  *                 ......
271  *                 `- indirect node ((6 + 2N) + x(N + 1))
272  *                       `- direct node
273  *                 ......
274  *                 `- indirect node ((6 + 2N) + (N - 1)(N + 1))
275  *                       `- direct node
276  */
277 static inline bool IS_DNODE(struct page *node_page)
278 {
279 	unsigned int ofs = ofs_of_node(node_page);
280 
281 	if (f2fs_has_xattr_block(ofs))
282 		return false;
283 
284 	if (ofs == 3 || ofs == 4 + NIDS_PER_BLOCK ||
285 			ofs == 5 + 2 * NIDS_PER_BLOCK)
286 		return false;
287 	if (ofs >= 6 + 2 * NIDS_PER_BLOCK) {
288 		ofs -= 6 + 2 * NIDS_PER_BLOCK;
289 		if (!((long int)ofs % (NIDS_PER_BLOCK + 1)))
290 			return false;
291 	}
292 	return true;
293 }
294 
295 static inline void set_nid(struct page *p, int off, nid_t nid, bool i)
296 {
297 	struct f2fs_node *rn = F2FS_NODE(p);
298 
299 	f2fs_wait_on_page_writeback(p, NODE);
300 
301 	if (i)
302 		rn->i.i_nid[off - NODE_DIR1_BLOCK] = cpu_to_le32(nid);
303 	else
304 		rn->in.nid[off] = cpu_to_le32(nid);
305 	set_page_dirty(p);
306 }
307 
308 static inline nid_t get_nid(struct page *p, int off, bool i)
309 {
310 	struct f2fs_node *rn = F2FS_NODE(p);
311 
312 	if (i)
313 		return le32_to_cpu(rn->i.i_nid[off - NODE_DIR1_BLOCK]);
314 	return le32_to_cpu(rn->in.nid[off]);
315 }
316 
317 /*
318  * Coldness identification:
319  *  - Mark cold files in f2fs_inode_info
320  *  - Mark cold node blocks in their node footer
321  *  - Mark cold data pages in page cache
322  */
323 static inline int is_file(struct inode *inode, int type)
324 {
325 	return F2FS_I(inode)->i_advise & type;
326 }
327 
328 static inline void set_file(struct inode *inode, int type)
329 {
330 	F2FS_I(inode)->i_advise |= type;
331 }
332 
333 static inline void clear_file(struct inode *inode, int type)
334 {
335 	F2FS_I(inode)->i_advise &= ~type;
336 }
337 
338 #define file_is_cold(inode)	is_file(inode, FADVISE_COLD_BIT)
339 #define file_wrong_pino(inode)	is_file(inode, FADVISE_LOST_PINO_BIT)
340 #define file_set_cold(inode)	set_file(inode, FADVISE_COLD_BIT)
341 #define file_lost_pino(inode)	set_file(inode, FADVISE_LOST_PINO_BIT)
342 #define file_clear_cold(inode)	clear_file(inode, FADVISE_COLD_BIT)
343 #define file_got_pino(inode)	clear_file(inode, FADVISE_LOST_PINO_BIT)
344 
345 static inline int is_cold_data(struct page *page)
346 {
347 	return PageChecked(page);
348 }
349 
350 static inline void set_cold_data(struct page *page)
351 {
352 	SetPageChecked(page);
353 }
354 
355 static inline void clear_cold_data(struct page *page)
356 {
357 	ClearPageChecked(page);
358 }
359 
360 static inline int is_node(struct page *page, int type)
361 {
362 	struct f2fs_node *rn = F2FS_NODE(page);
363 	return le32_to_cpu(rn->footer.flag) & (1 << type);
364 }
365 
366 #define is_cold_node(page)	is_node(page, COLD_BIT_SHIFT)
367 #define is_fsync_dnode(page)	is_node(page, FSYNC_BIT_SHIFT)
368 #define is_dent_dnode(page)	is_node(page, DENT_BIT_SHIFT)
369 
370 static inline void set_cold_node(struct inode *inode, struct page *page)
371 {
372 	struct f2fs_node *rn = F2FS_NODE(page);
373 	unsigned int flag = le32_to_cpu(rn->footer.flag);
374 
375 	if (S_ISDIR(inode->i_mode))
376 		flag &= ~(0x1 << COLD_BIT_SHIFT);
377 	else
378 		flag |= (0x1 << COLD_BIT_SHIFT);
379 	rn->footer.flag = cpu_to_le32(flag);
380 }
381 
382 static inline void set_mark(struct page *page, int mark, int type)
383 {
384 	struct f2fs_node *rn = F2FS_NODE(page);
385 	unsigned int flag = le32_to_cpu(rn->footer.flag);
386 	if (mark)
387 		flag |= (0x1 << type);
388 	else
389 		flag &= ~(0x1 << type);
390 	rn->footer.flag = cpu_to_le32(flag);
391 }
392 #define set_dentry_mark(page, mark)	set_mark(page, mark, DENT_BIT_SHIFT)
393 #define set_fsync_mark(page, mark)	set_mark(page, mark, FSYNC_BIT_SHIFT)
394